Function sort_and_categorize

deployments/feishu/daily-push-agent/main.py:1408–1468

排序并分类论文 Args: papers: 论文列表 profile: 用户画像 weights: 权重配置 Returns: 带分类的论文列表

(
    papers: List[Dict],
    profile: Dict,
    weights: Dict,
    progress_callback: Optional[Callable[[Dict[str, Any]], None]] = None,
)

def sort_and_categorize(
    papers: List[Dict],
    profile: Dict,
    weights: Dict,
    progress_callback: Optional[Callable[[Dict[str, Any]], None]] = None,
) -> List[PaperWithScore]:
    """
    排序并分类论文

    Args:
        papers: 论文列表
        profile: 用户画像
        weights: 权重配置

    Returns:
        带分类的论文列表
    """
    def emit(event: Dict[str, Any]) -> None:
        if not callable(progress_callback):
            return
        try:
            progress_callback(event)
        except Exception as exc:
            print(f"  Progress callback error: {exc}")

    result = []
    for paper in papers:
        score = calculate_paper_score(paper, profile, weights)
        relevance_signal = compute_relevance_signal(paper, profile)
        drift_bonus, drift_topics = compute_drift_bonus(paper, profile, weights)
        reading_signal_bonus, reading_signal_topics = compute_reading_signal_bonus(paper, profile, weights)
        score = min(1.0, score + drift_bonus + reading_signal_bonus)
        category = categorize_paper(score, paper, profile, weights)
        result.append(
            PaperWithScore(
                paper=paper,
                score=score,
                category=category,
                relevance_signal=relevance_signal,
                drift_bonus=drift_bonus,
                drift_topics=drift_topics,
                reading_signal_bonus=reading_signal_bonus,
                reading_signal_topics=reading_signal_topics,
            )
        )

    emit({"phase": "scored", "count": len(result)})

    # 先按分数粗排，再按分类结果做硬优先级重排
    result.sort(key=lambda x: x.score, reverse=True)

    result = categorize_papers_by_rank(result, profile, weights)
    result.sort(key=_hard_priority_tuple, reverse=True)
    result = apply_source_diversity_quota(result, weights)
    result = apply_mmr_topic_diversity(result, weights)
    result = apply_push_count_limit(result, weights)
    result.sort(key=_hard_priority_tuple, reverse=True)
    for rank, item in enumerate(result, start=1):